Mistral AI Launches Devstral 2

Mistral AI has launched Devstral 2, a new coding-focused AI model, to compete with other AI labs. The Devstral 2 series includes two open-source models: Devstral 2 (123B) and Devstral Small 2 (24B), designed for high-performance coding. Devstral 2 is a 123B-parameter dense transformer supporting a 256K context window.

Mistral has partnered with agent tools Kilo Code and Cline for the release of Devstral 2. Mistral is also working on Mistral Compute, an AI cloud. Devstral outperforms open-source models on SWE-Bench Verified by a large margin. It achieves a score of 46.8% on SWE-Bench Verified, exceeding prior open-source models by over 6% points.

Devstral is light enough to run on a single RTX 4090 or a Mac with 32GB RAM, making it suitable for local deployment and on-device use. It is available under the Apache 2.0 license, allowing for community building, customization, and acceleration of autonomous software development. The model is accessible on Mistral's API under the name devstral-small-2505, priced at $0.1/M input tokens and $0.3/M output tokens.